À propos de la FAO Actualités Multimédia Principaux sujets Statistiques Membres Publications

FAO/WHO GIFT | Global Individual Food Consumption Data Tool

Estimated usual intakes of selected nutrients

In the evaluation of population dietary intakes, one is often interested in the usual (habitual or long-term) intake of the population to describe the actual situation of the population with the adequacy of intakes.

An observed single-day of nutrient intake reflects two components of variation: the variation between persons and the variation within a person, the so-called day-to-day variation. The long-term average nutrient intake provides a better basis for understanding dietary intakes and further analysis, therefore the effect of day-to-day variation should be mitigated, to the extent possible. This can be achieved by applying statistical models. Adjusting for day-to-day variation results in a narrower distribution of intake, with the same mean intake, compared to the original data.

Available statistical models, developed to adjust dietary data for day-to-day variation, were reviewed to understand their potential application to dietary surveys shared through the FAO/WHO GIFT platform, together with their relative challenges, and limitations.

The statistical tool Statistical Program to Assess Dietary Exposure (SPADE), developed by the researchers from the National Institute for Public Health and the Environment (RIVM), The Netherlands, was selected to perform the analysis of usual intakes of nutrients (the usual intake analysis) for the FAO/WHO GIFT platform. Further refinement of the tool was undertaken to develop and generalize SPADE to allow for its application to datasets shared through FAO/WHO GIFT.

Information about SPADE and its outputs can be found in the following publications:

  • The usual intake analysis is only performed for datasets for which there are multiple non-consecutive days of 24hour recalls/food records for at least a subset of 50 individuals.
  • Datasets from surveys which were designed to capture differences in nutrient intakes across seasons for the same subject are considered not to be suitable for the usual intake analysis. This decision was supported by the fact that the results of the analysis did not fulfill the validation criteria (see section below) showing high within- to between-person variance ratios.
  • The usual intake analysis is performed independently from the data sharing process, hence for some eligible datasets that are available in the platform, the results for the usual intake analysis may not be available immediately after the data upload.
  • Due to SPADE’s requirements, subjects with missing sex are excluded from the usual intake analysis.
  • Due to SPADE’s requirements, subjects with missing age values are excluded from the usual intake analysis.
  • Due to SPADE’s requirements, age values are rounded down to the nearest integer.
  • Children aged less than 12 months are excluded from the usual intake analysis, due to a distinct method of estimating nutrient requirements and lack of Estimated Average Requirements (EARs) for nutrients for this age group.
  • The usual intake analyses are performed independently of the women’s physiological status, meaning that pregnant and lactating women are included in the total sample of women, without differentiation.
  • For dietary surveys including child-mother pairs, the results for the usual intake analysis are calculated separately for children and mothers, since the relationship between the consumption of both population groups would violate the independence assumption.

  • The usual intake analyses are performed using SPADE versions 4.1.16 and 4.1.26.
  • Usual intake distributions are estimated for selected nutrients based on the data availability in the original datasets.
  • For the currently selected nutrients, the 1-Part model is applied.
  • The usual intake analysis is performed in pre-defined population groups: females, males, and both females and males, in the following main and sub age groups, designed to allow for gradual disaggregation of data by age:
    • 1 to 8 years old
           - 1 to 3 years old
           - 4 to 8 years old
    • 9 to 18 years old
          -  9 to 13 years old
           -14 to 18 years old
    • 14 years and older
          - 14 to 50 years old
          - 51 years and older
    • 19 years and older
          - 19 to 50 years old
          - 51 years and older
  • Normalized survey weights or the population data for the year of data collection are incorporated in the analysis, if available for the survey.
  • The minimum sample size for the calculation and presentation of the usual intake analysis results is 50 subjects. If there are less than 50 subjects in the aforementioned population groups, the results are not displayed.

The statistical procedure applied by SPADE to estimate usual intake distributions can be divided into three main steps:

  1. Box-Cox transformation of the nutrient intake data to obtain approximatively normal intake distributions.
  2. Using transformed data, model the nutrient intakes as a function of age on the transformed scale and estimate .  both the within-person and between-person variability. This allows to generate a sample of pseudo persons (representing the population) with their mean usual intake in the transformed scale.
  3. Back-transformation of the mean usual intakes of (II) results in a sample of habitual intakes at the original scale, which can be described by percentiles (per age unit (years) or custom age groups), to ensure a meaningful presentation of results for the population.

Additionally, quantification of uncertainty is provided by bootstrapped confidence intervals of all percentiles and the mean of the usual intake distribution.

  • For each dataset analyzed, the applied statistical model is validated by:
    • Assessment of Box-Cox transformation effectiveness: whether values of lambda are between -0.1 and 1.0.
    • Assessment of the within- to between-person variance ratio: whether the ratio is between 0.25 and 5.
    • Analysis of regression residuals, e.g. residual plots and standard deviation of residuals: whether they follow a Gaussian distribution.
  • If, during the model validation stage, the model fails the test of the within- to between-person variance, the model is corrected by removing outliers. Outliers are identified by computing squared differences between the individual observations across two recalls and observing the largest gap between consecutively ordered squared differences. Those subjects which fall beyond the observed gap are iteratively considered outliers and removed from the sample. In the interest of preserving the maximum number of observations in the dataset, this corrective action is applied only to the population groups for which the model has a within- to between-person variance ratio larger than 5.
  • Additional validation steps include:
    • Assessment of the mean estimated usual intake and the lower and upper bounds: whether they fall into plausible intervals and whether the mean estimated usual intakes are greater than their lower bound and lower than their higher bound.
    • Assessment of the percentiles of the estimated distributions: whether they fall into plausible intervals and whether the values are increasing.
    • Assessment of the median (P50) estimated usual intake and the lower and upper bounds: whether they fall into plausible intervals and whether the median is greater than the lower bound and lower than the upper bound.

The results of the usual intake analysis using SPADE are presented as a function of age on the percentile scale and can be directly compared with Nutrient Reference Values (NRVs) to estimate the percentage of a population at risk of not meeting their nutrient requirements. However, the results of such comparison should be interpreted with caution given the limitations specific to each dataset, the available NRVs, as well as the uncertainty in the estimated usual intake distribution, which is also displayed on the platform.

The FAO/WHO GIFT platform provides information about the distribution of estimated usual intakes of selected nutrients among pre-defined population groups, without comparing the results to the NRVs or providing interpretation in terms of nutrient adequacy. To facilitate this assessment by users, the platform provides easy access to NRVs compiled from three sources: Report of a joint FAO/WHO Expert Consultation on Human Vitamin and Mineral Requirements (FAO/WHO)1,2 scientific opinions on dietary reference values published by the European Food Safety Authority (EFSA)3,4,5,6,7,8,9 and dietary reference intakes published by Institute of Medicine (IOM)10. Following the recommended approach for the evaluation of inadequate nutrient intakes among population groups11,12, Estimated Average Requirements (EAR) and equivalent NRVs representing the nutrient requirements of 50% of healthy individuals, such as Adequate Intake (AI) and Average Requirement (AR) are presented on the dashboard.

 

 

1 World Health Organization. (‎2006)‎. Guidelines on food fortification with micronutrients / edited by Lindsay Allen … [‎et al.]‎. World Health Organization. https://apps.who.int/iris/handle/10665/43412

2 World Health Organization. (‎2004)‎. Vitamin and mineral requirements in human nutrition, 2nd ed. World Health Organization. https://apps.who.int/iris/handle/10665/42716

3 EFSA NDA Panel (EFSA Panel on Dietetic Products, Nutrition and Allergies), 2013a. Scientific Opinion on Dietary Reference Values for vitamin C. EFSA Journal 2013;11(11):3418. doi:10.2903/j.efsa.2013.3418

4 EFSA NDA Panel. 2013b. Scientific Opinion on nutrient requirements and dietary intakes of infants and young children in the European Union. EFSA Journal 2013;11(10):3408 doi:10.2903/j.efsa.2013.3408

5 EFSA NDA Panel. 2015a. Scientific Opinion on Dietary Reference Values for vitamin A. EFSA Journal 2015;13(3):4028 doi:10.2903/j.efsa.2015.4028

6 EFSA NDA Panel. 2015b. Scientific Opinion on Dietary Reference Values for iron. EFSA Journal 2015;13(10):4254. doi:10.2903/j.efsa.2015.4254

7 EFSA NDA Panel. 2015c. Scientific opinion on Dietary Reference Values for Calcium. EFSA Journal 2015;13(5):4101. doi:10.2903/j.efsa.2015.4101

8 EFSA NDA Panel. 2014. Scientific opinion on Dietary Reference Values for zinc EFSA Journal 2014;12(10):3844. doi:10.2903/j.efsa.2014.3844

9 EFSA NDA Panel. 2015d. Scientific opinion on Dietary Reference Values for folate EFSA Journal 2014;12(11):3893. doi:10.2903/j.efsa.2014.3893

10 HMD (Health and Medicine Division of the USA National Academies of Sciences). 2016. Dietary Reference Intakes Tables and Application - Estimated Average Requirements and Adequate Intakes.  http://www.nationalacademies.org/hmd/Activities/Nutrition/SummaryDRIs/DRI-Tables.aspx

11 Lindsay H Allen, Alicia L Carriquiry, Suzanne P Murphy, Perspective: Proposed Harmonized Nutrient Reference Values for Populations, Advances in Nutrition, Volume 11, Issue 3, 2020, Pages 469-483, ISSN 2161-8313, https://doi.org/10.1093/advances/nmz096.

12 Principles of Nutritional Assessment – 2nd ed., Rosalind S. Gibson, Oxford University Press, 2005

< Back